Collusion via Signaling in Multiple Object Auctions with Complementarities- An Experimental Test*

We experimentally study bidder collusion in open ascending auctions for multiple objects. The project is based on the theoretical results by Brusco and Lopomo (1999), who give theoretical support for the following claims: (1) simultaneous ascending bid auctions can be vulnerable to collusion in the multi-object case; (2) The sole presence of complementarities does not hinder collusion; (3) Collusion is a “lown umbers” phenomenon. We focus on a simultaneous ascending auction for two objects. Several experimental treatments are considered: markets with low numbers (2 bidders) and high numbers (5 bidders), no complementarities (additive values) and complementarities (superadditive values). Experimental results are largely consistent with the theory. Collusion is often observed in two-person markets with or without complementarities. Previous experience under the same treatment greatly facilitates bidder collusion. There is no evidence of collusion in five-person markets. We further study collusive strategies adopted by bidders in two-person markets. While most strategies make extensive use of signaling, in the presence of complementarities, bidders use collusive strategies that are supported only by repeated play.